DeepRefiner

DeepRefiner refines protein structures using deep neural network architectures to improve the accuracy of predicted atomic models.

Key Features:

• High-accuracy refinement: Performs refinement of protein structures to improve model accuracy and reliability.
• Deep neural network architectures: Employs advanced deep neural network architectures calibrated for structure refinement.
• Flexible refinement modes: Provides 'adventurous' and 'conservative' modes to modulate the degree and consistency of adjustments.
• CASP validation: Demonstrated performance in CASP13 and CASP14 under the group name 'Bhattacharya-Server'.

Scientific Applications:

• Protein model refinement: Improves atomic-level accuracy of predicted protein structures for structural biology and modeling studies.
• Benchmarking in CASP: Participated in CASP13 and CASP14, ranking No. 2 among refinement servers (second only to 'Seok-server' in CASP13 and 'FEIG-S' in CASP14).

Methodology:

Uses advanced deep learning techniques based on deep neural network architectures calibrated to provide adjustable ('adventurous' or 'conservative') refinement modes.